Mars Technology Program
Reliable and Efficient Long-Range
Autonomous Rover Navigation
Introduction
Upcoming planetary rovers will embark on ambitious campaigns of exploration. The Mobile Science Laboratory scenario calls for a highly capable new class of rover to explore distinct science sites separated by a distance of several kilometers, collecting vast amounts of scientific data.
Important Mars science can be accomplished through detailed examination of constrained science sites but planetary surface investigation will move beyond detailed examination to regional exploration. Regional exploration is motivated by different science objectives, focusing on geologic units, their distributions and properties, and the discovery of interesting features within or at the contact of these units in order to understand the geologic record and development of the planet. Regional exploration is made possible by reliable long-range autonomous navigation.
This project addresses Rover Technology in the area of Long-Range Autonomous Navigation. We are maturing this technology to NASA Technology Readiness Level 6 (TRL-6), through experimentation in a relevant environment.
Using our solar-powered rovers Hyperion and Zoe we have demonstrated preliminary results in long-range autonomous rover navigation with traverses of 300-600 meters in a single command cycle using 30 meter resolution digital terrain maps and onboard odometric information (meaning without GPS). We achieved one instance of single-command 1-kilometer autonomous rover traverse in the Atacama Desert, April 2003. This project will enable planetary rovers to routinely exceed 1 kilometer in single-command autonomous traverse.
Hyperion on long traverse in the Chilean Atacama Desert, April 2003
We are collaborating with the NASA ASTEP project "Life in the Atacama" to experimentally verify our approach with Zoe in field experiments planned and funded for 2005. The synergy of these projects will serve to improve the rover capability and its performance in life detection for ASTEP while achieving quantitative analysis of our approach for the MTP program.
Objectives and Approach
Our objective is to enable long traverses that are reliable and efficient. By reliable, we mean that the rover minimizes the risk that it will be lost, damaged, or require remote operation. By efficient, we mean that the rover minimizes critical resources to accomplish its objectives.
We are taking a three-pronged approach:
- Continuous Cost Field Path Planning: Funded under the previous NASA MTP contract, TEMPEST/ISE is the most advanced path planner/re-planner for planetary rovers. The planner is capable of optimizing some metrics while satisfying others. In practice, it has been used to generate the shortest paths between way points for a solar-powered rover without exhausting the battery, and to re-plan when the rover falls behind schedule or new information about the terrain is discovered. But TEMPEST/ISE is limited to discretized map and planning space representations, and the paths are not truly optimal. We are extending TEMPEST/ISE to recover much of that lost optimality while still satisfying all constraints, thus improving both efficiency and reliability.
- Learned Map/Sensor Data Association: Current map-based planners revise their maps in a rote sense: only the portion measured by the robot's sensors is updated. This way of using map data is common to many autonomous navigation programs, including NASA's "Life in the Atacama" project and DARPA's PerceptOR program. For this project, we will learn associations between sensor and map data in order to apply updates more globally, thus maximizing the benefit of sensor measurements for terrains that are only known coarsely. The newly planned routes will be based on a more accurate map, and thus will be more reliable and efficient.
- Far-Field Perception and Terrain Modeling: Present navigation algorithms focus on the near-field, leaving a gap between the maximum range analyzed by the rover and the maximum resolution of orbital models. This myopic view is prone to driving into rover-scale obstacles like embankments, drainages, micro-craters, hummocks, and dunes. We are extending the perception analysis to the far-field and modeling to multiple scales to enable planning for these terrain features.
We are continuing our collaboration with the JPL CLARAty architecture [Volpe, et al, 01] development begun in the JPL Mars Exploration Technology program. In this project we are continuing to create CLARAty-compatible software to establish capabilities for long-distance rover navigation. A significant contribution of this project is to develop the needed navigation and planning components using CLARAty and to test using a CLARAty-compliant rover.